← Back to Architecture
QKV fusion
ArchitectureUsed in
1 PRs
Best BPB
1.1171
Avg BPB
1.1171
Submissions
Hyperparameters Across PRs
| pr_number | parameters |
|---|---|
| 670 | {"q_heads":8,"kv_heads":4} |
| pr_number | parameters |
|---|---|
| 670 | {"q_heads":8,"kv_heads":4} |